Procedural Decisions in Intergroup Conflict

Transcription

1 Social Justice Research, Vol. 9, No. 4, 1996 Procedural Decisions in Intergroup Conflict Situations: The Influence of Hierarchical Group Membership Kees van den Bos,1,2 Riel Vermont,' and Henk A. M. Wilke' A review of research on procedural and distributive justice shows that whereas distributive justice research has examined people's actual choice of outcomes, procedural justice research has paid little attention to the investigation of people's actual choice of procedures. In the present paper; three experiments are presented, all investigating people's actual choice of procedures. In all three experiments, participants were in a middle-management position where their subordinates demanded an opportunity to voice their opinion, while their superiors demanded that the subordinates should not be allowed voice. In Experiments 1 and 2 it was found that participants who were induced to identifr with the low hierarchical group (subordinates) allowed more voice than participants who were induced to identifr with the high hierarchical group (superiors), but that the effect of hierarchical group membership was absent when maximizing performance (Experiment 1) or participative values (Experiment 2) were explicitly emphasized. In Experiment 3 it was found that the effect of hierarchical group membership on procedural decision making was also evident in persons who progressed from identifying with both hierarchical groups to identifring with one hierarchical group. In the discussion it is argued that cross-fertilization between the fields of procedural and distributive justice will deepen our understanding of social justice in general. KEY WORDS: procedural justice; group membership; procedural decisions; organizational hierarchy. People may distinguish the fairness of outcomes that are allocated from the fairness of the way in which those outcomes are arrived at. In other 'Department of Social and Organizational Psychology, Leiden University. P.O. Box 9555, 2300 RB Leiden, The Netherlands. 2A11 correspondence should be sent to Kees van den Bos /96/ / Plenum Publishing Corporation

2 372 van den Bos, Vermunt, and Wilke words, people may make a distinction between issues of distributive and procedural justice (Lind and 'Tyler, 1988: Tyler and Lind, 1992). For example, in organizations a manager may have to decide about the fair allocation of outcomes (e.g., how many tasks her subordinates should complete), or she may have to decide about the fair procedure with which those outcomes are arrived at (e.g., whether she will allow her subordinates an opportunity to voice their opinions about how many tasks they have to complete: cf. Lind et al., 1990). As Greenberg (1987) has noted, people may also distinguish between reactive and proactive justice. That is, people may be motivated to escape from, or avoid, unfair situations (a reactive justice issue), or to actively secure, or approach, fair situations (a proactive justice issue). For example, subordinates may react to distributive or procedural decisions of their manager by indicating that they judge these decisions to be more or less fair (e.g., they may indicate that they have been requested to complete an excessive number of tasks or that they should be allowed some voice in the process of deciding how many tasks they should complete). The manager, on the other hand, may want to actively secure a fair situation by deciding that her subordinates should complete a fair number of tasks or by deciding that her subordinates should be allowed some fair amount of voice in the distributive decision process. Thus, when conducting social justice research, we may distinguish between two dichotomies underlying social justice issues. Furthermore, as Greenberg (1987) has pointed out, these two dichotomies can be combined, resulting in a 2 x 2 taxonomy. The main advantage of this taxonomy is that it highlights the interrelationships between the four orientations, and in so doing identifies parallel types of research that have not yet been conducted. Most notably, the taxonomy revealsas was also concluded by Greenberg (1987)that whereas proactive distributive research has examined directly how people make distributive decisions, proactive procedural research has not paid much attention to the investigation of how people actually make procedural decisions. Prototypical research showing the proactive distributive justice orientation is an experiment by Leventhal and Michaels (1969). In this experiment participants were asked to allocate money between themselves and a confederate participant. The amount of money participants allocated constituted the main dependent variable. The limited amount of research on proactive procedural justice has been of three types. In the first type of study, participants are asked to list examples of fair and unfair procedures (e.g., Lissak and Sheppard, 1983, Expt. 1). In the second type of study, participants rate the importance of various procedures presented in written scenarios (e.g., Lissak and Shep-

3 Procedural Decisions 373 pard, 1983, Expt. 2). Of particular relevance to how procedural decisions are actually made is the third type of study, in which people's preferences for various procedures (Thibaut et al., 1974; Azzi, 1993) or people's reported use of various procedures are investigated (Lind et al, 1994). Thus, whereas participants in proactive distributive research actually make a distributive decision (e.g., allocate money), participants in the proactive procedural research that we know of at best only indicate their opinion about which procedural decision the experimenter should make or report which procedures they have actually used. However, a behavioral opinion or reported behavior does not necessarily predict actual behavior in an accurate way. Therefore, theory development in the proactive procedural justice domain might profit from research that observes directly how people make procedural decisions. The first objective of the present paper is to present such research. In the field of procedural justice one of the most significant theories is Lind and Tyler's group-value model (1988; cf. Tyler and Lind, 1992), which is based on Tajfel and Turner's (1979) social identity theory and Turner's (1985) self-categorization theory. On the basis of these two theories, Tyler and Lind (1992, pp ) have argued that a group advocates procedures that reflect its fundamental values. For example, a group that allows group members to voice their opinions reflects a different attitude with regard to group members than a group that does not allow such voice. Thus, the procedure a group advocates is viewed as a manifestation of fundamental group values. As argued by Tyler and Lind on the issue of reactive procedural justice, this implies that when a group advocates a fair procedure towards one of its members, the group is indicating a positive relationship to that group member. We argue, however, that on the issue of proactive procedural justice this implies thatbecause group members usually want a positive relationship with their group (Tajfel and Turner, 1979) group members choose the procedure they view as a manifestation of the fundamental values of their group. Tyler and Lind (1992) have remarked that most group-value research has focused "on the search for information about one's position within one's group, rather than on the position of one's ingroup vis-à-vis other groups" (p. 143). However, as argued in social identity theory and self-categorization theory, what constitutes a group depends not only on the perception of intragroup similarities but also of intergroup differences. In other words, a group cannot be defined in isolation from other groups. The above analysis of the literature on social justice suggests that it may be useful to investigate procedural decisions in intergroup situations (i.e., to explore the implications for procedural decisions of being a member of a group that advocates a particular procedure as compared with being a member of a

4 374 van den Bos, Vermont, and Wilke group that advocates another procedure). This is an issue that has only recently received attention in group-value research (Lind et al., 1994), in contrast with research on social identity theory and self-categorization theory. The earliest studies on procedural justice were laboratory experiments (e.g., Thibaut et al., 1974), which were criticized for their artificial character (see Lind and Tyler, 1988, for an overview). In the present paper, therefore, we sought to conduct laboratory experiments that simulate important reallife situations. A remarkable example of a situation where procedural decisions are made under conditions of conflicting procedural expectations of different groups is that faced by middle managers (cf. Mintzberg, 1983). In organizations, a middle manager is often faced with two groups of people: a group of superiors, which is hierarchically above the middle manager, and a group of subordinates, which is hierarchically below the middle manager. Furthermore, these two groups may be expected to favor different procedures. Prior research (cf. Mintzberg, 1983) suggests that low hierarchical group members feel that they are entitled to demand voice in distributive decision processes that are relevant to them, whereas high hierarchical group members feel entitled to demand that the low hierarchical group should not be allowed voice. In the present study we used the middle-management situation to investigate procedural decision making. As mentioned above, from the group-value model it may be assumed that group members choose the procedure they view as a manifestation of the fundamental values of their group. Thus, in a situation where a middle manager is a member of a hierarchical group that values a particular procedure and another hierarchical group values a different procedure, the middle manager may choose the procedure that is advocated by his or her group, rather than the procedure advocated by the other group. However, we argue here that the middle-management situation is also very interesting since it emphasizes that people can belong to several groups at the same time (whichin contrast to social identity and self-categorization theoryis an issue that has not been studied in group-value research thus far, and hence may help to extend the group-value model): The middle manager may be a member not only of a (low or high) hierarchical subgroup but is simultaneously a member of the organization as a whole. Furthermore, from the organizational literature (e.g., Mintzberg, 1983) we know that organizations may have particular values that are viewed as fundamental to that organization. For instance, an organization may emphasize that all organizational members should perform well and may evaluate its members accordingly. Other organizations, on the other hand, are known for their emphasis on participation in decision making as a fundamental value. From our above analysis of the group-value model it may be assumed that when

5 Procedural Decisions 375 an organization as a whole explicitly emphasizes its fundamental values, these organizational valuesbeing more general than hierarchical subgroup valuesmay affect a middle manager's procedural decisions more strongly than the values the middle manager's hierarchical subgroup advocates. Therefore, in addition to the effects of hierarchical group membership, the influence of organizational values on procedural decisions is investigated in the present paper. EXPERIMENT 1 In Experiment 1, participants are placed in a hierarchical position between a superior and three subordinates. A distributive decision is to be made: It is to be decided how many tasks the subordinates should complete. The task of the participants is to make the procedural decision: to decide how long (0-10 min) the subordinates will be allowed to express their opinions about the number of tasks that they have to complete (cf. Lind et al., 1990). Participants are furthermore told that the subordinates (i.e., the recipients of the procedure) want a considerable amount of time to voice their opinion. The superior, however, demands that the subordinates should not be allowed voice. Because the earliest procedural justice experiments (e.g., Thibaut et al., 1974) were criticized for their artificial character, we made the experimental situation more realistic by informing participants that the number of minutes that voice would be allowed would be subtracted from the available work time that the subordinates would be allowed to work on the tasks. Participants are informed they are either a member of the high hierarchical group (superiors) or the low hierarchical group (subordinates). On the basis of the group-value model we propose that participants will favor the procedure advocated by the group with which they identify. Thus, participants who identify with the low hierarchical group (which demands a considerable amount of voice) are expected to allow more voice than participants who identify with the high hierarchical group (which demands that no voice should be allowed). However, as was mentioned above, on the basis of the group-value model it may also be assumed that there may be other sources of influence on procedural decisions besides the hierarchical group membership of the decision maker. More specifically, we argue thatsince organizational values are more general than hierarchical subgroup valuesit can be derived from the group-value model that when an organization as a whole favors the values of one of its hierarchical groups, all organizational members will conform more to the values of that hierarchical group. Empirical evidence

6 376 van den Bos, Vermunt, and Wilke (e.g., Mintzberg, 1983) suggests that organizations tend to emphasize productivity more than the well-being of subordinates. In Experiment 1, therefore, half of the participants are explicitly faced with the demand that participants should perform well, whereas the other half are not faced with such an explicit performance demand. On the basis of the group-value model it is assumed that when performance is emphasized all participants might conform more to the procedure advocated by the high hierarchical group (i.e., no voice). Thus, Hypothesis 1: When participants are explicitly faced with the demand that all participants should perform well they will allow less voice than when they are not explicitly faced with such a performance demand. However, whereas emphasis on optimal performance by the organization might lead all participants to favor the procedure advocated by the high hierarchical group, when the superordinate goal is more ambiguous participants may allow more voice when in a low as opposed to a high hierarchical group. Thus, Hypothesis 2: Participants who are not explicitly faced with a performance demand might allow more voice when in a low hierarchical group than when in a high hierarchical group, but that this hierarchical group membership effect on procedural decision making may be less strong when participants are explicitly faced with a performance demand. Method Participants and Design Eighty students (35 men, 45 women) at Leiden University participated in the experiment and were paid 7.50 Dutch guilders for their participation. Participants were randomly assigned to one of the conditions of the 2 (Performance Demand: absent, present) x 2 (Hierarchical Group: high, low) factorial design. Experimental Procedure Participants were invited to the laboratory to participate in a study on how people work in computer networks. Upon arrival at the laboratory, participants were led to separate cubicles, each containing a computer with a monitor and a keyboard. Participants were led to believe that a minimum of six participants were present. Furthermore, participants were told that the computers were connected to one another, and that the experimenter could communicate with them by means of the computer network. The

7 1 Procedural Decisions 377 computers were used to present stimulus information, and to collect data on the manipulation checks and the dependent variable. At the start of the experiment, the participants in the Performance Demand Present condition were informed that all participants would be evaluated on how well they performed their tasks. Participants in the Performance Demand Absent condition were only informed that all participants would be evaluated on how well they functioned in general. In both Performance Demand conditions, participants were told that the participant with the best evaluation would earn a cake. (In fact, after all participants were run, a cake was randomly given to one participant, a procedure to which none of the participants objected.) After this, the participants were told that they would work in a fiveperson group. The persons in the group were labeled A, B, C, D, and E. Participants were told that each participant would be assigned a different position. Actually, all participants were assigned to position B. Participants in the High Hierarchical Group condition were told that A and B constituted the high hierarchical group of leaders, and that C, D, and E constituted the low hierarchical group of work performers. Furthermore, these participants were informed that the position of A was comparable to that of a senior manager, and that the position of B was comparable to that of a middle manager. Participants were also told that As position was slightly higher than B's position. The positions of C, D, and E were said to occupy a single hierarchical level comparable to that of work performers. Participants in the Low Hierarchical Group condition were told that A constituted the high hierarchical group of leaders, and that B, C, D, and E constituted the low hierarchical group of work performers. The position of A was said to be comparable to that of a senior manager. The position of B was described as comparable to that of a foreman supervising work performers C, D, and E. Therefore, B's position was said to be slightly higher than those of C, D, and E. After this, the tasks of A, B, C, D, and E were explained. Since As position was comparable to that of a senior manager, As task was to supervise B, C, D, and E. The task of C, D, and E was described as a figure detection task: Figures would be presented on the right part of the computer screen. Each figure consisted of 36 squares, and each square showed one of eight distinct patterns. On the left part of the computer screen one of the eight patterns would be presented, and the task of C, D, and E was to count in the figure on the right part of the screen the number of squares with this pattern. After this C, D, and E, would perform the figure detection task for 15 min. Furthermore, participants were informed that the number of figures C, D, and E should count had to be determined. This was the distributive

8 378 van den Bos, Vermunt, and Wilke decision. The task of B was to decide how many minutes C, D, and E would be allowed to voice their opinions about the number of figures they should complete. This was the procedural decision. Participants could allow C, D, and E an amount of voice time between a minimum of 0 min and a maximum of 10 min. Participants were presented with fictitious survey results indicating that C, D, and E held the opinion that they should be allowed the maximum amount of voice (i.e., 10 min voice time), and that A held the opinion that B should not allow any voice to C, D, and E (i.e., 0 min voice time). To make the experimental situation more realistic, participants were presented with the justifications A, C, D, and E had presumably given for their opinions, and were informed that the number of minutes that voice was allowed would be subtracted from the available work time that C, D, and E were allowed to count the figures (in a pilot study without these justifications and time subtraction, we obtained similar results). Participants were also instructed that their task was to make the fairest procedural decision (see TOrnblom, 1992). After receiving information about how many figures C, D, and E had counted in a practice round, participants made the procedural decision. They were told that their procedural decision would be communicated to C, D, and E. At this point participants responded to questions pertaining to the manipulation checks. When the participants had completed these questions, the experiment ended, and participants were paid for their participation and thoroughly debriefed. Results and Discussion Manipulation Checks Participants were asked to indicate the criterion on which all participants were to be evaluated (1 = how well they performed; 2 = how well they functioned in general). Logit analysis produced a model with the twoway association between Performance Demand and this manipulation check, and the first-order effect of the manipulation check. The model showed adequate fit between observed and expected frequencies, likelihood ratio G2(2) = 2.82, p = Ninety-eight percent of the Performance Demand Present participants answered that participants would be evaluated on how well they performed, and 98% of the Performance Demand Absent participants answered that participants would be evaluated on how well they functioned in general.

9 1 Procedural Decisions 379 To check social identification, participants were asked to indicate on a 7-point Likert type scale whether they agreed with the statement that they identified with the group of leaders (1 = strongly disagree; 7 = strongly agree). A 2 (Performance Demand) x 2 (Hierarchical Group) analysis of variance (ANOVA) yielded a main effect of Hierarchical Group only, F(1, 76) = 79.99, p < High Hierarchical Group participants (1 = 6.1) identified more with the group of leaders than Low Hierarchical Group participants (1 = 3.8). Participants were also asked whether they identified with the group of work performers. An ANOVA yielded a main effect of Hierarchical Group only, F(1, 76) = 47.86, p < Low Hierarchical Group participants ( Y = 4.9) identified more with the group of work performers than High Hierarchical Group participants ( x- = 2.7). It can be concluded that the independent variables were perceived as intended. Procedural Decisions The dependent variable was the procedural decision participants made. Participants were asked "How many minutes (0-10) voice time do you allocate to C, D, and E?" As predicted by Hypothesis 1, a 2 x 2 ANOVA yielded a main effect of Performance Demand, F(1, 76) = 8.97, p < 0.01, showing that Performance Demand Absent participants allowed more voice (Tc = 4.7) than Performance Demand Present participants (1 = 3.8). Furthermore, and as predicted by Hypothesis 2, the ANOVA also yielded a significant interaction, F(1, 76) = 4.20, p < Inspection of the means revealed (Fig. 1) that whereas in the Performance Demand Absent condition Low Hierarchical Group participants allowed more voice (I = 5.3) than High Hierarchical Group participants (5i = 4.1), t(38) = 2.82, p < 0.01, this effect of hierarchical group membership was absent in the Performance Demand Present condition (Y = 3.8), t(38) = -0.21, ns. It can be concluded that the findings of Experiment 1 support our predictions derived from the group-value model, in that it appears that when the organization as a whole emphasizes work productionand by implication shows less concern for voicemiddle managers may allow less voice than when the organization is more ambiguous about its values. Moreover, when the organization as a whole emphasizes work production, middle managers' procedural decisions may not be influenced by the hierarchical group with which middle managers identify. If, however, the organization is ambiguous about its values, middle managers who identify

10 I 380 van den Bos, Vermont, and Wilke 6 5 Procedural Decisions _ Performance Demand Absent A Performance Demand Present 4 - A A 3 Low High Hierarchical Group, Fig. 1. Procedural decisions (0-10 min of voice granted to subordinates) as a function of Hierarchical Group and Performance DemandExperiment 1. with the group of subordinates may allow more voice than middle managers who identify with the group of superiors. The findings of Experiment 1 suggest that when an organization as a whole emphasizes the values of a high hierarchical group of superiors, middle managers will conform to the values of that hierarchical group. An interesting implication of the findings of Experiment 1 is that a similar pattern of results might be found when an organization that emphasizes the values of a low hierarchal group of subordinates is compared with an organization that is more ambiguous about its dominant value: Middle managers may allow more voice to subordinates in organizations favoring participative values than in organizations that are ambiguous about their values (Hypothesis 1). Moreover, because emphasis on participative values by the organization might lead all participants to favor the procedure advocated by the low hierarchical group, it is hypothesized that participants who are not explicitly faced with a participative demand will allow more voice when in a low hierarchical group than when in a high hierarchical group, but that this effect of hierarchical group membership on procedural decision making will be less strong when participants are explicitly faced

11 Procedural Decisions 381 with a participative demand (Hypothesis 2). This implication is investigated in Experiment 2. EXPERIMENT 2 Method Participants and Design Eighty students (22 men, 58 women) at Leiden University participated in the experiment. They were paid 10 Dutch guilders for their participation and were randomly assigned to one of the conditions of the 2 (Participative Demand: absent, present) x 2 (Hierarchical Group: high, low) factorial design. Experimental Procedure The experimental procedure of Experiment 2 was similar to that of Experiment 1: At the beginning of the experiment, the participants in the Participative Demand Present condition were informed that they would be evaluated on how well they paid attention to the wishes of subordinates. Participants in the Participative Demand Absent condition were told that they would be evaluated on how well they functioned in general. In both Participative Demand conditions, participants were told that the participant with the best evaluation would earn a cake. (Actually, after all participants were run, a cake was randomly given to one participant, a procedure to which none of the participants objected.) The experimental set-up was similar to that of Experiment 1: The participants were told that they would work in a six-person group. The persons in the group were labeled A, B, C, D, E, and E Participants were instructed that A, B, and C constituted the high hierarchical group of leaders, and that D, E, and F constituted the low hierarchical group of work performers. Furthermore, it was communicated that the position of A was comparable to that of a top manager, that B's position was comparable to that of a senior manager, and that C's position was comparable to that of a middle manager. Participants were also informed that As position was slightly higher than B's position, and that B's position was slightly higher than C's position. Also, D's position was said to be comparable to that of a foreman supervising the senior work performer E and the junior work performer F, that D's position was slightly

12 382 van den Bos, Vermunt, and Wilke higher than E's position, and that E's position was slightly higher than F's position. After this, the tasks of A, B, C, D, E, and F were explained: It was communicated that As task was comparable to that of a top manager, B's task was comparable to that of a senior manager, C's task was comparable to that of a middle manager, D's task was comparable to that of a foreman, E's task was comparable to that of a senior work performer, and F's task was comparable to that of a junior work performer. At this point, positions were assigned. Participants in the High Hierarchical Group condition were assigned to position C, and participants in the Low Hierarchical Group condition were assigned to position D. Furthermore, participants in the Low Hierarchical Group condition were informed that, because they had been assigned to position D, they would receive specific information about the tasks of E and E Participants in the High Hierarchical Group condition were told that, because it had already been explained that D's task was comparable to that of a foreman, they would receive specific information about the tasks of E and E Participants were then informed that the task of E and F was to perform the figure detection task for 15 min. Furthermore, participants were informed about the distributive and procedural decisions, and that it was their task to make the procedural decision: to allow E and F voice time for some duration between 0 and 10 min. Participants were then presented with fictitious survey results: In the High Hierarchical Group condition, the survey results indicated that D, E, and F held the opinion that E and F should be allowed the maximum amount of voice (10 min voice time), and that A and B held the opinion that C should not allow any voice to E and F (0 min voice time). In the Low Hierarchical Group condition, the survey results indicated that E and F held the opinion that they should be allowed 10 min voice time, and that A, B, and C held the opinion that D should allow 0 min voice time to E and E lb make the experimental situation more realistic, and in accordance with Experiment 1, participants were presented with justifications which other participants allegedly had given for their opinion, on voice time, and were informed that the number of minutes that voice was allowed would be subtracted from the available work time that E and F were allowed to count the figures. Participants were also instructed that their task was to make the most fair procedural decision. After receiving information on how many figures E and F had counted in a practice round of 15 min, participants made the procedural decision. Participants were told that their procedural decision would be communicated to E and E Participants then responded to questions pertaining to the manipulation checks. When the participants had completed

13 Procedural Decisions 383 these questions, the experiment ended, and participants were paid for their participation and thoroughly debriefed. Results and Discussion Manipulation Checks To check the induction of Participative Demand, participants were asked whether they agreed with the statement that they would be evaluated on how well they paid attention to the wishes of the subordinates. A 2 (Participative Demand) x 2 (Hierarchical Group) ANOVA yielded a main effect of Participative Demand only, F(1, 72) = 22.57, p < Participants in the Participative Demand Present condition ( i = 5.9) agreed more that they would be evaluated on how well they paid attention to the wishes of the subordinates than participants in the Participative Demand Absent condition (1 = 3.7). Participants were also asked whether they agreed with the statement that they would be evaluated or how well they functioned in general. An ANOVA yielded a main effect of Participative Demand only, F(1, 72) = , p < Participants in the Participative Demand Absent condition ( f = 6.7) agreed more that they would be evaluated on how well they functioned in general than participants in the Participative Demand Present condition (7 = 3.0). To check social identification, participants were asked whether they identified with the group of leaders. An ANOVA yielded a main effect of Hierarchical Group only, F(1, 72) = , p < High Hierarchical Group participants (7 = 6.8) identified more with the group of leaders than Low Hierarchical Group participants (7 = 1.9). Participants were also asked whether they identified with the group of work performers. An ANOVA yielded a main effect of Hierarchical Group only, F(1, 72) = , p < Low Hierarchical Group participants (7 = 6.6) identified more with the group of work performers than High Hierarchical Group participants (I = 1.5). It can be concluded that the independent variables were perceived as intended. Procedural Decisions Participants were asked: "How many minutes (0-10) voice time do you allocate to E and F?" As predicted by Hypothesis 1, a 2 x 2 ANOVA yielded a main effect of Participative Demand, F(1, 72) = 22.78,p < 0.001, showing that participants in the Participative Demand Present condition

14 384 van den Bos, Vermunt, and Wilke allowed more voice (Y = 6.7) than participants in the Participative Demand Absent condition (I = 4.8). Furthermore, as predicted by Hypothesis 2, the ANOVA also yielded a significant interaction, F(1, 72) = 4.33, p < Inspection of the means revealed (Fig. 2) that in the Participative Demand Absent condition Low Hierarchical Group participants allowed more voice (.7 = 5.4) than High Hierarchical Group participants (7 = 4.1), t(38) = 2.76, p < As expected, this effect of hierarchical group membership was absent in the Participative Demand Present condition (I = 6.7), t(38) = -0.59, ns. The findings of Experiments 1 and 2 support the group-value model: When an organization as a whole emphasizes performance demands or participative values (and by implication shows respectively less or more concern for voice) middle managers may allow respectively less or more voice than when the organization is more ambiguous about its values. Moreover, when an organization favors performance values or participative values, middle managers' procedural decisions may not be influenced by the hierarchical group with which middle managers identify. However, if an organization is more ambiguous about its values, middle managers who Procedural Decisions *Participative Demand Absent *Participative Demand Present Low High Hierarchical Group Fig. 2. Procedural decisions (0-10 min of voice granted to subordinates) as a function of Hierarchical Group and Participative DemandExperiment 2.

15 Procedural Decisions 385 identify with the group of subordinates may allow, more voice than middle managers who identify with the group of superiors. EXPERIMENT 3 The findings of Experiments 1 and 2 show that an effect of hierarchical group membership on procedural decision making is evident between persons who identify with different hierarchical groups. However, on the basis of Rawls's (1971) justice-as-fairness theory it may be argued that an important question remains whether the influence of hierarchical group membership will be evident within persons who progress from being unaware to being aware of their hierarchical group membership. In his contract theory about the construction of a just society, Rawls makes a distinction between procedural decisions made when one is not aware of the position one will occupy in society (one is then "behind a veil of ignorance") versus procedural decisions made when one is aware of one's position (one is then "before the veil of ignorance"). Furthermore, Rawls argues that a person who does not know the position he or she will occupy in society will make the most fair procedural decision; that is, this noncategorized person will make a decision that is in the interest of the most needy Ṫhe distinction between procedural decisions made behind or before a veil of ignorance has been investigated in social justice research (e.g., Azzi, 1993: Thibaut et al., 1974). An often ignored aspect of Rawls' theory, however, is that after people have made a procedural decision behind a veil of ignorance they are expected to return to society and become categorized. People usually make procedural decisions when they occupy a place in society, and therefore we focus in the present paper on this ignored aspect of Rawls's theory. To investigate this issue, two phases are introduced in Experiment 3. In Phase 1, participants are not informed to which hierarchical group they will be assigned. As in Experiments 1 and 2, participants are told that subordinates demand that they should be allowed 10 min voice time and that the superior demands that participants should allow 0 min voice time. Then participants are asked to make a procedural decision. In Phase 2, participants are actually assigned to either the high or the low hierarchical group, after which they are asked to make a (second) procedural decision. From the group-value model it may be derived that in Phase 1 participants will identify as much with the high hierarchical group as with the low hierarchical group, whereas in Phase 2 they will shift their social identification to either the high hierarchical group or the low hierarchical group

16 386 van den Bos, Vermunt, and Wilke (depending on the specific hierarchical group they are assigned to). As a consequence, it is hypothesized that in Phase 1 participants will pay attention to the procedures benefiting both groups, whereas in Phase 2 they will shift their attention, paying more attention to the procedure their group advocates than to the procedure the other group advocates. Thus, it is hypothesized that participants who are assigned to the high hierarchical group in Phase 2 will allow less voice in Phase 2 than in Phase 1 (Hypothesis 1), and that participants who are assigned to the low hierarchical group in Phase 2 will allow more voice in Phase 2 than in Phase 1 (Hypothesis 2). However, with regard to Hypothesis 2 an alternative prediction may be derived from the justice as fairness theory (Rawls, 1971): As was argued above, on the basis of Rawls theory it may be assumed that persons who are not aware of the position to which they will be assigned will select the alternative that is in the interest of the most needy (i.e., the subordinates). For our experimental situation this implies that after participants have been assigned to the low hierarchical group, they might make the same procedural decision as that which they made when they were uninformed about their group membership. In other words, for the participants who are assigned in Phase 2 to the low hierarchical group, Rawls's theory predicts no shift in procedural decision making (Hyptothesis 2alt). Method Participants and Design Forty students (26 men, 14 women) at Leiden University participated in the experiment. They were paid 10 Dutch guilders for their participation. The experiment utilized a 2 (Hierarchical Group: high, low) x 2 (Phase: phase 1, phase 2) factorial design. Hierarchical Group was manipulated between subjects, Phase was manipulated within subjects, and participants were randomly assigned to one of the Hierarchical Group conditions. Experimental Procedure The procedure of Experiment 3 was similar to the procedures of Experiments 1 and 2: First of all, participants were told that they would work in a six-person group consisting of positions labeled A, B, C, D, E, and E A and B constituted the high hierarchical group of leaders, and C, D, E, and F constituted the low hierarchical group of work performers. As position was comparable to that of a senior manager, and B's position was comparable to that of a middle manager. It was also communicated to the

17 Procedural Decisions 387 participants that the position of A was slightly higher than B's position. C's position was described as comparable to that of a foreman supervising the work performers D, E, and F, and C's position was said to be slightly higher than the positions of D, E, and E In Phase 1, it was communicated to the participants that they would be assigned to either position B or position C, but they were not informed of which position they would be assigned to. Participants were then informed that A supervised all of the other participants, and that C supervised D, E, and F The task of D, E, and F was to perform the figure detection task for 15 min. Furthermore, participants were informed about the distributive and procedural decisions, and that it was their task to make the procedural decision. On the basis of a fictitious survey, participants were led to believe that D, E, and F held the opinion that they should be allowed the maximum amount of voice (10 min voice time), and that A held the opinion that no voice should be allowed (0 min voice time). Furthermore, to make the experimental situation more realistic, and in accordance with Experiments 1 and 2, participants were presented with justifications which the participants allegedly had given for their opinions on voice time, and were informed that the number of minutes that voice was allowed would be subtracted from the available work time that E and F were allowed to count the figures. Participants were also instructed that their task was to make the most fair procedural decision. Then participants were asked to make the procedural decision. After making the procedural decision, participants answered questions included as manipulation checks for Phase 1. In Phase 2 participants were actually assigned to their position. Participants in the High Hierarchical Group condition were assigned to position B, and participants in the Low Hierarchical Group condition were assigned to position C. After this, participants were instructed to make a second procedural decision, and to make the most fair procedural decision. When participants had made the procedural decision, they responded to questions included as manipulation checks for Phase 2. After this the experiment ended, and participants were thoroughly debriefed and paid for their participation. Results and Discussion Manipulation Checks To check social identification, participants were asked in both Phase 1 and Phase 2 whether they agreed with the statement that they identified

18 1 388 van den Bos, Vermunt, and Wilke with the group of leaders. A 2 (Group Membership) x 2 (Phase) ANOVA with repeated measures on the last factor yielded a significant interaction, F(1, 38) = 79.41,p < 0.001, as well as a main effect of Hierarchical Group, F(1, 38) = 60.22, p < 0.001: Participants in the High Hierarchical Group condition identified more with the group of leaders in the second phase (I = 6.6) than in the first phase (X = 4.6), t(19) = -6.32, p < Participants in the Low Hierarchical Group condition identified less with the group of leaders in the second phase (X = 2.2) than in the first phase ( 17 = 4.5), t(19) = 6.32, p <.001. Participants were also asked in both Phase 1 and Phase 2 whether they identified with the group of work performers. An ANOVA yielded a significant interaction, F(1, 38) = 76.42, p < 0.001, and a main effect of Group Membership, F(1, 38) = 48.79, p < 0.001: In the Low Hierarchical Group condition, participants identified more with the group of work performers in the second phase (X = 6.0) than in the first phase (X = 3.2); t(19) = -6.76, p < In the High Hierarchical Group condition, participants identified less with the group of work performers in the second phase (X = 1.6) than in the first phase (X = 3.4), t(19) = 5.55, p < It can be concluded that the independent variables were perceived as intended. Procedural Decisions Participants made a procedural decision twice: In both Phase 1 and Phase 2 participants were asked: "How many minutes (0-10) voice time do you allocate to D, E, and F?" A 2 (Group Membership) x 2 (Phase) ANOVA with repeated measures on the last factor yielded a significant interaction only, F(1, 38) = 9.53, p < Inspection of the means (see Fig. 3) indicated some evidence for Hypothesis 1: Participants in the High Hierarchical Group condition allowed less voice in the second phase (I = 3.9) than in the first phase (X = 4.1), although this effect failed to reach significance, t(19) = 1.56, p < As predicted by Hypothesis 2, and in contrast with Hypothesis 2,ilt, participants in the Low Hierarchical Group condition allowed more voice in the second phase (X = 4.5) than in the first phase (X = 3.9), t(19) = -2.67, p < The findings of Experiment 3 suggest that participants who were uninformed in Phase 1 and were assigned to the high hierarchical group in Phase 2 allowed less voice in the second phase than in the first phase. This effect, however, failed to reach significance (i.e., p < 0.07). How can we explain this nonsignificance? One explanation for the nonsignificance may be derived by taking into account that for these participants, a decrease in

19 Procedural Decisions 389 5, Procedural Decisions -6- Low Hierarchical Group -A- High Hierarchica Group 4 3 Phase 1 Phase Phase 2 Fig. 3. Procedural decisions (0-10 min of voice granted to subordinates) as a function of Phase and Hierarchical GroupExperiment 3. voice was involved: They were expected to give less voice in the second phase than in Phase 1 (in contrast with the participants who were assigned to the low hierarchical group in Phase 2, who were expected to give more voice in the second phase than in the first phase). As Törnblom (e.g., 1992) has argued with respect to distributive decision making, giving less of a desired resource can be considered a negative allocation whereas giving more of a desired resource can be considered as a positive allocation. Furthermore, the work of Törnblom suggests that people consider the allocation of negative outcomes to be more difficult than the allocation of positive outcomes, and hence are more reluctant to make negative distributive decisions than to make positive distributive decisions. Our findings seem to generalize Tornblom's point to the domain of procedural decision making: People seem more reluctant to shift to giving less voice than to shift to giving more voice. The findings of Experiment 3 do show that participants who were uninformed in Phase 1 and were assigned to the low hierarchical group in Phase 2 allowed more voice in Phase 2 than in Phase 1 (when they identified as much with the high hierarchical group as with the low hierarchical

20 390 van den Bos, Vermont, and Wilke group). Thus, whereas Rawls's (1971) theory would predict that these participants should have made the same procedural decision in the second phasewhen they were assigned to the low hierarchical group as in the first phasewhen they were uninformed about their group membership our findings indicate that these participants showed a shift in procedural decision making. While it must be admitted that our experimental manipulationsas well as the experimental manipulations of Azzi (1993) and Thibaut et al (1974)did not create in full the veil of ignorance as described by Rawls (1971), our findings seem to disconfirm the prediction we derived from Rawls that participants would choose the same procedure when they were behind a veil of ignorance as when they were assigned to the disadvantaged group. GENERAL DISCUSSION In accordance with the group-value model, the present findings suggest that procedural decisions made by people in such situations of intergroup conflict are influenced by the hierarchical group with which the procedural decision makers identify: Procedural decision makers who identify with a low hierarchical group that demands that subordinates should be allowed a considerable amount of voice allow more voice than procedural decision makers who identify with a high hierarchical group that demands that no voice should be allowed. Furthermore, the effect of social identification on procedural decision making is evident not only between persons who identify with different hierarchical groups (Experiments 1 and 2) but also within persons who progress from identifying with both hierarchical groups to identifying with one hierarchical group (Experiment 3). The group-value model also suggests that when an organization as a whole favors the values of one hierarchical group, middle managers conform more to the values of that hierarchical group than when the organization does not explicitly emphasize the values of one hierarchical group. The findings of Experiments 1 and 2 suggest that when the organization as a whole emphasizes the values of one hierarchical group, procedural decision makers pay more attention to the procedure advocated by that hierarchical group. Furthermore, in such situations of accentuated group values, the effect of hierarchical identification on procedural decision making was less powerful. This suggests that the group-values of an organization as a whole may affect procedural choice. It should be noted here that we have employed strict definitions of procedure and outcome. For instance, departing from the definition of a procedure as the way an outcome is achieved, a procedural decision was

21 Procedural Decisions 391 defined as a decision about which procedure will be used to yield an outcome. In many contexts, there are clear differences between procedures and outcomes (see Tyler and Lind, 1992), but in some situations the difference between a procedure and an outcome is more ambiguous. With regard to the present research, for example, it remains to be seen whether we would have obtained similar effects of hierarchical group membership if participants had made not a procedural decision, but a distributive decision (e.g., allocating money). Theoretically, this may imply that the four justice orientations are less distinct than is suggested by Greenberg's (1987) taxonomy. As this may be the case, the present paper, as one of the first attempts to investigate proactive procedural justice directly, may serve as a starting point for future research on proactive procedural justice. In proactive distributive justice research several studies have been conducted investigating under what circumstances people will allocate outcomes according to an equity rule or an equality rule. This suggests a need for proactive procedural justice research to investigate how people decide between two procedural rules. For example, suppose participants are brought into a situation in which they will have to evaluate the performance of their subordinates. Furthermore, suppose that, due to time constraints, participants are forced to decide which of two procedures they will apply: The procedure used to make the performance evaluation may either be very accurate (i.e., all the relevant information known about subordinates' performance is very carefully checked) or allow subordinates an opportunity to voice their opinion about the performance evaluations; the participants cannot select both an accurate procedure and a voice procedure, however (due to the time constraints). Future research might explore the circumstances under which participants are likely to choose the voice procedure as opposed to the accurate procedure, and vice versa. The present paper may also offer useful suggestions for research on proactive distributive justice. For example, the research presented in this paper suggests that it may be useful to explore the influence of hierarchical group membership on people's distributive decisions (e.g., the allocation of money). Furthermore, reactive distributive justice may profit from the insights provided by procedural justice theory. For instance, procedural justice research suggests that not only instrumental but also noninstrumental (or symbolic) concerns may affect people's procedural fairness judgments to a great extent. It might be argued, therefore, that distributive fairness judgments also may be affected by symbolic concerns. For example, in general waiters may prefer to receive a tip as opposed to receiving no tip. However, suppose someone gives the waiter a tip of 5 cents. In our opinion,